The first thing you notice on Duvo.ai's homepage is a sentence that reads less like marketing and more like a confession: "The documented process and the real process are never the same." [Duvo.ai] Below it, a quiet product story unfolds about agents that watch the actual clicks, the spreadsheet workarounds, the supplier portal logins that nobody put in the SOP, and then quietly take them over. There is no astronaut imagery, no glowing brain. For a category that usually shouts, the restraint reads as a thesis: retail operations are not glamorous, and the software that fixes them should not pretend otherwise.
That thesis now has $15 million behind it. In December, Duvo.ai closed a seed round led by Index Ventures, with Northzone and Credo Ventures joining [Finsmes, Dec 2025]. The Prague-based company, founded in 2025, is selling what it calls an AI workforce for retail and FMCG operations: agents that monitor carrier portals, parse delivery exceptions, push price updates across ERP and ecommerce systems, and clear the queue of repetitive work that currently sits on the desks of category managers and operations analysts [Duvo Blog]. According to the company, early deployments are running inside multi-billion-dollar retail and FMCG groups on six-figure annual contracts [Retail Technology Innovation Hub, Dec 2025].
The bet
Duvo's wedge is speed of deployment into a software environment that is famously hostile to it. Retail operations live in SAP, in supplier portals, in carrier dashboards, in legacy pricing systems, and in the Excel files that bridge all of them. The company's pitch is that its agents can be live in weeks rather than the multi-quarter implementations that characterize traditional process automation [Duvo Blog]. Pricing is structured as a fixed price per execution rather than a token meter, which the company argues makes the ROI legible to a CFO on day one [Duvo Blog].
The product positioning takes direct aim at SAP Build Process Automation, the incumbent low-code workflow tool that ships with the SAP estate. Duvo's own writing on the comparison concedes that SAP Build is a "legitimate product with real capabilities," then argues its boundaries break down the moment a workflow needs to cross outside the SAP ecosystem into a supplier portal or a marketplace [Duvo Blog]. That is the seam Duvo is trying to pry open: the integration tax, the human API problem, the analyst whose actual job is to copy numbers from one tab to another.
Why it could be big
The tailwind here is not subtle. European grocery and FMCG operators are running on margin profiles that make every percentage point of operating cost matter, and the back-office headcount that supports pricing, replenishment, and supplier coordination has scaled linearly with SKU complexity for two decades. An agent layer that can absorb even a fraction of that work has obvious appeal, and the investor syndicate Duvo assembled reflects it. Index Ventures has a long history of backing European software, Northzone backed Spotify and Klarna, and Credo Ventures is one of the most active early-stage funds in Central and Eastern Europe.
Seed funding raised | 15 | $M
Reported team size | 15 | people
Company-cited manual work reduction | 40 | %
The round size itself, $15 million at seed, is unusually large for a company founded the same year, and signals that the investors are underwriting a specific person more than a specific traction curve.
The team and traction
That person is Tomas Čupr, co-founder and CEO, who previously founded Rohlik, the European online grocery business that became one of the region's best-known consumer unicorns [The Recursive]. He is joined by co-founders Martin Pecha and Marek Paris [Seedtable], and the broader founding group is described as industry operators with decades of retail experience [Duvo Blog]. The team currently sits at roughly 15 people [Forbes]. Čupr has written publicly that the motivation for Duvo grew out of watching how much of a modern retailer's day is spent moving data between systems that were never designed to talk to each other, and wanting to build the AI workforce that gives that time back [tomascupr.substack.com].
The customer signal is the most interesting part of the early story. Six-figure annual contracts inside multi-billion-dollar retail and FMCG groups [Retail Technology Innovation Hub, Dec 2025] is not the shape of a pilot program. It suggests Duvo is being bought by operators who already know what an integration tax feels like, likely through Čupr's own network from the Rohlik years.
The honest counterfactual
What bears will say is that enterprise retail automation is a graveyard of ambitious wedges, and that SAP, despite its limitations, has a structural advantage: it is already in the building. SAP Build Process Automation is bundled into commercial relationships that procurement teams renew without thinking [Duvo Blog]. Any agent layer that lives outside the SAP estate has to win the politics as well as the technology, and the sales cycle for a six-figure contract inside a global FMCG can stretch a year. What bulls answer is that Duvo's pricing model and deployment speed are designed precisely to short-circuit that cycle, and that the company is selling into the gap SAP itself acknowledges, the work that crosses ecosystem walls. The early six-figure contracts suggest that argument is landing in at least a handful of buying committees.
What to watch
The next twelve months will tell us whether Duvo's agent layer generalizes. Watch for the first publicly named customer logo, which would convert the Rohlik halo into independent proof. Watch for the team to roughly double from its current 15, particularly on the forward-deployed engineering side, which is the constraint on weeks-not-years deployment promises. And watch for a Series A in late 2026 if the six-figure contracts compound into seven-figure ones; with Index already on the cap table, the path to a much larger round is short if the renewal data cooperates.
The deeper question Duvo is implicitly answering is one the consumer-facing internet has been chewing on for a year now, just relocated to the back office: when an agent can do the work, what exactly is the job?